Party organisation and the party-delegate style of representation

Representational roles

Representational roles refer to the way in which representatives conceive of their role in the democratic system (Önnudóttir and von Schoultz, 2021). It is an attitudinal concept, relating most directly to the general set of values representatives hold and their socialisation to norms. Since the seminal work of Eulau et al. (1959), the literature has mainly discussed two aspects of representational roles (Blomgren and Rozenberg, 2012; Esaiasson, 2000; Eulau and Karps, 1977; Wahlke et al., 1962): first, the focus of representation, referring to the group(s) in society the representative considers herself to represent (e.g. her constituency, the nation as a whole, or specific groups of voters); and second, the style of representation: ‘what criterion the representative ought to use when making decisions’ (Önnudóttir and von Schoultz, 2021: 122). Three styles of representation are usually distinguished according to the ‘source’ or ‘principal’ that guides the representative’s decision (Converse and Pierce, 1986): the party-delegate style, in which representatives believe above all in following their party’s line, the voter-delegate style, according to which representatives follow their (constituency) voters’ preferences, and the trustee style, following which representatives take their decisions independently, without reference to any external principal. An additional category of ‘politicos’ describes those who alternate between a trustee and a delegate style depending on the context (Wahlke et al., 1962) or whose choices do not clearly indicate a predominant style. Styles of representation have implications for the focus of representation: party-delegates believe that they primarily represent their party, which may in turn represent certain groups of voters; voter-delegates represent their constituency voters; and trustees are more inclined to conceive of themselves as representatives of the national interest.

In this study, we focus on styles of representation, and being a party-delegate (rather than a voter-delegate, a trustee, or a politico) in particular. The party-delegate style of representation is important, as it underpins central assumptions of the Responsible Party Model (Pierce, 1999).

It is separate from, but has direct implications for representatives’ behaviour (Andeweg, 2012) - including voting behaviour during plenary sessions (see e.g. Carey and Shugart, 1995 and (Shomer, 2009; Shomer, 2016) for studies of MPs’ behaviour and their determinants). The concept of party ‘loyalty’ has also been used in the literature addressing the issues of party cohesion and/or party discipline (Andeweg and Thomassen, 2011; Close and Gherghina, 2019). However, the concept of ‘loyalty’ can be confusing: it is sometimes used to refer to representatives’ attitudes towards their party, sometimes to their (voting) behaviour (for a broader discussion, see Close and Gherghina 2019). In this contribution, we clearly want to address representatives’ attitudes towards their party, by relying on theoretical concepts developed in the field of representation studies. We believe that the representation literature offers the best theoretical framework to analyse these attitudes, as it allows to reflect upon the role of both values and interests in shaping attitudes towards representation.

Empirically, we will examine the influence of party attributes on the percentage of candidates in each party adopting the party-delegate style; thus, we examine party-delegate outcomes at party level. This party-level approach follows Önnudóttir (2016: 734) who argues that the collective, party level is vital to representation. The contention that the mode of representation at party level matters is also complemented by the party organisation literature in which organisational types are associated with different representational styles (e.g. Katz and Mair, 1995).

Explaining the styles of representation

Two trends can be discerned among the scientific works dedicated to explaining representational styles. The first of these trends is theoretical. Two theoretical perspectives have dominated the literature, and have sometimes been combined: approaches rooted in rational choice theories have conceived of representational roles as strategic responses to specific incentives (Searing, 1994; Strøm, 1997), while representational roles have also been conceived of as ‘habits of the heart’ resulting from socialisation processes (Zittel, 2012). The second trend is empirical: most attention has been devoted to macro-level and micro-level determinants of representational roles, with less attention directed towards the influence of the meso- or party-level factors.

By conceiving of candidates and legislators as mainly driven by their prospects of (re)election (Mayhew, 1974), rational choice approaches have found that systems in which the election of individuals depends more on party reputation than on their personal reputation (Carey and Shugart, 1995) produce more representatives who consider themselves party-delegates. In this regard, two variables have been examined: first, district magnitude, with larger districts associated with party-centred systems and smaller or single-member district systems associated with an emphasis on candidates’ personal reputations (Farrell and Scully, 2010; Önnudóttir & von Schoultz, 2021; Pilet et al., 2012); and second, the degree of openness of the ballot, distinguishing between systems allowing preferential voting and closed-list systems, with the latter encouraging a party-delegate style (Önnudóttir & von Schoultz, 2021; Wessels and Giebler, 2011). Individuals’ perceptions of their chances of winning their district (Zittel, 2012) or being elected (Wessels and Giebler, 2011) have also been examined; findings suggest that secure candidates are less in hock to their party.

Previous studies have also highlighted the explanatory role of individual-level determinants, mostly related to political experience within the party (e.g. duration of party membership or political experience within the organisation) and experience in local or national elected assemblies (Önnudóttir, 2016; Önnudóttir and von Schoultz, 2021; Sudulich et al., 2020; Wessels and Giebler, 2011; Zittel, 2012). Experience increases individuals’ tendency to adopt a party-delegate style as a result of socialisation within the party and public offices (Önnudóttir & von Schoultz, 2021; Önnudóttir, 2016; Sudulich et al., 2020). Socialisation effects could also be related to candidates’ and legislators’ sociodemographic traits, such as age and gender. While the effect of age is not straightforward (Kam, 2009: 199; Sudulich et al., 2020), some evidence seems to suggest that women are less likely to ‘rebel’ attitudinally against the party (Cowley and Childs, 2003; Close, 2018: 213), and consider themselves more dependent on external authorities (parties or voters) when making their decision (Önnudóttir and von Schoultz, 2021; Sudulich et al., 2020).

Existing studies that have explored the role of parties and their organisations have mostly relied on rational choice approaches and have conceived of party characteristics as institutional constraints. In this vein, a common argument is that when the selection of candidates is controlled by the party leader, candidates tend to adopt a party-delegate style (Esaiasson, 2000; Önnudóttir, 2016; Strøm, 2012; Wessels and Giebler, 2011): party leaders reward loyal members, and they strategically nominate politicians who tend to stick to the party line. Other party level variables include party size in the parliamentary arena (Gherghina, 2011) or in terms of vote share in the constituency (Wessels and Giebler, 2011: 12) and government incumbency (Önnudóttir, 2016; Sudulich et al., 2020), but the findings in respect of these variables are quite inconsistent across studies.

Building on socialisation approaches, some scholars have considered party ideology or party family. However, the findings are limited, and the theory is quite underdeveloped. According to some accounts, parties of the right should be more likely to have a higher number of trustees, while parties of the left should have more ‘partisans’ or party-delegates (Andeweg, 1997; Önnudóttir, 2016; Wessels and Giebler, 2011). Damgaard’s (1997) and Heidar’s (1997) studies slightly nuance this left-right divide by uncovering party differences that might be related to parties’ culture or ideology. Showing that party-delegates dominate both the two main parties in Norway, Heidar (1997: 106) further argues that the party is probably no longer as central in shaping representational roles as the Duvergerian tradition suggests. Sudulich et al. (2020) and Önnudóttir and von Schoulz (2021) find that Green party candidates are more likely than other candidates to consider themselves trustees, and thus less likely to consider themselves as being subject to an outside authority (party or voter); they also find this to be the case for Liberal parties’ candidates. Close (2018) argues that party family differences are due to different ‘norms’ of party solidarity, and different sets of values (e.g. individualism vs. authoritarianism) embodied in the parties’ ideologies, but also in their organisational structure and ‘culture’. Nevertheless, these analyses do not distinguish between two correlates of party family: ideology and organisation. In this contribution, we aim to contribute to better understand the role of organisational characteristics in shaping candidates’ attitudes.

Political parties and styles of representation: hypotheses

Our first hypothesis relates to a party’s organisational strength, with an emphasis on financial or human resources that it can deploy in support of its candidates. The idea that party organisational strength influences representational styles has its roots in the seminal works of Converse and Pierce (1986) and of Wahlke et al. (1962), which suggested that the difference between the dominant trustee style of representation in the US presidential system and the dominant party-delegate style of European parliamentary party-based systems result from the weakness of party organisations in the US compared to their relative strength in Europe. Yet, this organisational weakness has been inadequately operationalized in further studies on representational styles; at best, scholars have used proxies such as party size and incumbency in executive office (see above).

Some research on party unity adopts a more varied and organisation-focused operationalisation of party strength. Little and Farrell (2017) investigate the influence of party strength on legislative party unity, and they measure party strength through party income and membership (both relative to the size of the electorate), as well as government incumbency and party size in parliament. In her study of parties’ legislative voting unity in post-Communist countries, Tavits (2012) demonstrates a positive relationship between party strength (operationalised as membership, local organisational strength, and electoral strength) and legislative unity. Although she does not consider party income as a measure as such of organisational strength, her statistical models control for party budget (as a percentage of the country’s GDP) and uncover a positive relationship with party unity.

The line of argument concerning party organisational strength that is pursued by Tavits (2012) and Little and Farrell (2017) provides the basis for our first hypothesis. Party strength in the form of resources is relevant in the sense described above, which is consistent with the rational choice perspective: party resources provide incentives for representatives to be more party-oriented in their attitudes, as a well-resourced party has greater potential to help them to achieve their goals, and to hinder them by withholding resources. We consider two types of organisational resources – financial (party income) and human (party membership) – combining them into a single measure of organisational strength (details below).

H1: The greater is the party’s resource-strength, the higher the proportion of party-delegates.

Our second set of hypotheses aims to test how intra-party power relations can shape the prevalence of the party-delegate style. Existing research focuses on the effect of candidate selection in terms of (de)centralisation. The argument is mostly rooted in a rational choice perspective and identifies party discipline as the core mechanism: when the nomination of candidates rests in the hands of the leadership (a centralised or exclusive process), candidates comply with the party line and adopt a party-delegate style since being loyal to the party (leader) would be a rewarding strategy for MPs seeking re-selection (Esaiasson, 2000; Önnudóttir, 2016). Also from a rational choice point of view, leaders who can control the selection procedures are expected to be more likely to nominate candidates who value party loyalty (Sieberer, 2006). They could also use more centralised procedures as punitive means to discipline the party (Ceron, 2016). These arguments, however, concern the legislators’ behaviour responding rationally to organisational incentives and produce inconclusive results when the questions of attitudes and role perception are introduced (e.g. Close et al., 2019). We could however assume that the introduction of stronger party discipline will not only have the direct effect of changing representatives’ behaviour but will also contribute to diffusing a culture that is more likely to shape their attitudes towards the party. Hence, a ‘centralisation’ hypothesis can be formulated as following:

H2: If the party leader has the final say in the selection of candidates, then the proportion of party-delegates is higher.

Yet, the argument proposed by Gauja (2012) provides another interpretation of this phenomenon at work. She suggests that the party’s organisational culture, resulting from the party’s origin, shapes the way members of the parliamentary party relate to the rest of the party organisation. Hence, grassroots parties born outside of parliament would have developed an organisation in which the parliamentary party is seen as the delegate of the members and the wider organisational structure, resulting in a party-delegate style of representation. By contrast, parties formed in parliament would tend to contain more trustees, as the organisation puts a greater emphasis on ‘the independence of parliamentarians in their organisational ethos’ (Gauja, 2012: 122). This echoes this idea that mass parties would embody a much stronger orientation by representatives towards their large number of extra-parliamentary members and their somewhat homogenous voters (Katz and Mair, 1995).

We extend this line of reasoning, although we do not conceptualise intra-party culture by referring to ‘types’ of party organisation. Rather, we rely on indicators of the degree of membership-orientation of party organisations, understood as the extent of grassroots influence on party activities and decision-making. We aim to measure party members’ influence within the organisation by delineating their role in candidate selection processes, which should directly affect candidates (hypothesis 3), and their role in the broader intra-party structure, which may have a less direct effect (hypothesis 4).

We argue in hypothesis 3 and 4 that power accorded to members within the organisation will grant the party-on-the-ground more controlling prerogatives over the activities of party representatives, and stronger norms of representatives being accountable to the membership and, thus, the party. We first examine the effect of members’ power to select candidates on the proportion of party-delegates within parties. This ‘decentralisation’ hypothesis states that where the selection of candidates rests ultimately in the hands of party members, parties will have more party-delegates as a result of rational incentives that this presents to candidates, but also because of the organisational culture with which it is associated. This runs contrary to the conventional argument that party leaders, rather than members, select party-delegates, and therefore is in tension with hypothesis 2. What both leader-centred and membership-centred forms of organisation are likely to have in common, however, is a stronger culture of the party compared to parties in which other faces of the party organisation have the final say, which in turns shapes attitudes towards representation.

H3: If party members have the final say in the selection of candidates, then the proportion of party-delegates is higher.

Second, we consider in the same vein the power of party members in the broader party structure, excluding their role in candidate selection. We understand ‘grassroots influence’ in the sense of the assembly-based type of intra-party democracy (Poguntke et al., 2016; von dem Berge and Poguntke, 2017), which allows members to participate (directly or via delegates) within the organisation at different stages of the processes. From a sociological point of view, candidates’ conception of their representational roles in internally democratic parties will reflect the internalisation of the norm of party solidarity. As a result, MPs and candidates in membership-oriented organisations are more likely to identify primarily as a delegate of the broader party organisation. Hence, we expect that:

H4: The more democratic are intra-party structures, the higher the proportion of party-delegates.

Data sources

To test these hypotheses, we rely on an overlapping sample from two cross-national projects: the Comparative Candidates Survey ¹⁶ and the PPDB ¹ (Poguntke et al., 2016). Our dataset includes 5825 candidates in 11 national parliamentary elections that took place between 2013 and 2017 (see Table 2 in the Appendix for details). In total, 70 parties are covered by our analyses, although there is some missing data for a few parties (detailed below), which reduces the number of parties to 60 in the final models. Each party includes between 13 and 275 respondents (candidates), with a median of 60.5 respondents per party.

Measuring candidates’ style of representation

As outlined above, three styles of representation have been extensively discussed in the literature: the trustee style, the party-delegate style, and the voter-delegate style. These styles have often been measured empirically using survey items that ask legislators or candidates about how they should vote in case of a difference between their own opinion, that of their party, and that of the voters. The voters’ opinion is the most difficult to apprehend, and may refer to the parties’ voters or to the candidate’s constituency voters. Similarly, we use three questions from the CCS, each implying a choice between two ‘principals’:

D4. How should a Member of Parliament vote in parliament?

D4a: If the voters in his/her constituency have one opinion and his/her party takes a different opinion

  1 = According to party’s opinion

Four categories of candidate can be distinguished empirically: party-delegates (candidates choosing the party in items D4a and D4c), trustees (candidates choosing their own opinion in items D4b and D4c), voter-delegates (candidates choosing his/her voters in items D4a and D4b), and candidates who do not consistently choose one principal when presented with it as part of different combinations (the politicos). The distribution of candidates across these categories is as follows: 1808 (31%) respondents can be categorized as party-delegates, 1315 (22.5%) as voter-delegates, 2235 (38.5%) as trustees and 467 (8%) as being part of the politico category. While each category represents a substantial proportion of the candidates in the sample, the size of the politicos category also suggests that most respondents have rather clear preferences and opinions about the style of representation they want to apply. This distribution is in line with previous studies (Önnudóttir, 2016; Wessels and Giebler, 2011).

In the analysis, we test our hypotheses by examining variation in the percentage of party-delegates by party, following the approach of Önnudottir (2016). While existing analyses often examine individual-level variations, party-level variations have somehow been overlooked. Our analysis precisely makes a contribution by analysing variations across parties and party “types” (in terms of organisation, family, size etc.). This party-level focus allows to consider each party as equal to another – unsensitive to the number of respondents by party.

The percentage of party-delegates by party varies from a minimum of 4.3% (ANEL in Greece) to a maximum of 77.8% (the radical right SD in Sweden), with an average of 28.9% and a median at 25.3% (see Table 2 in the Appendix for details by party and for a list of all party acronyms). There is considerable variation by country, with the median Norwegian and Swedish parties having the highest proportions of party-delegates (>40%), followed by Belgium and Italy (>30%), and then by other countries (<25%). There is also some variation by party family, with the median Green parties and Liberal parties (<20%) having lower proportions of party-delegates than other party families.

Party strength: financial resources and party membership

Two types of intra-party resources are considered in relation with the first hypothesis, from which a measure of party organisational strength is derived using factor analysis. Party income is expressed in (or converted to) Euro. We standardize it by the size of the country’s electorate (thus, income per member of the electorate), then by the national economy (income per member of the electorate as a proportion of GDP) (Poguntke et al., 2016; van Biezen and Kopecky, 2015: 5). This variable ranges from a minimum of 0.0000917 (i.e. the German AfD) to a maximum of 0.058 (i.e. the radical right Norwegian Progress Party), with an average of 0.004 and a standard deviation of 0.008. Data is missing for three parties. ²

The second intra-party resource, party membership, is measured through the use of the Membership/Electorate (M/E) ratio has the merit of showing the proportion of the electorate that are members of a party in a given system. This indicator also makes it possible to compare membership rates between parties in the same country over time and in different countries (Kölln, 2014: 76; Mair and Van Biezen, 2001; Van Haute and Gauja, 2015). The variable ranges from 0.0000971 (the Hungarian Greens, LMP) to 0.031 (Belgian PS). ³ High scores are found mainly in former mass parties – such as the Norwegian Labour Party, the Swedish Social Democrats or the Flemish Christian Democrats (CD&V). Note that data is missing for seven parties. ⁴

These two variables do not follow a normal distribution: they are in fact highly skewed. We will therefore use their log in the regression models. Note that when we log the two variables and then examine their correlation, the correlation at the party level is around 0.49 (p-value < .001). One way to deal with this collinearity ⁵ is to reduce the information to one dimension. For that purpose, we use a factor analysis that results in one component ⁶ (eigenvalue = 1.49), accounting for about 75% of the variation. Cronbach’s alpha (0.66) indicates that this scale is quite reliable. In the analysis, we will use this party strength component to account for both resources at the same time.

Intra-party power: candidate selection and intra-party democratic structure

In order to test H2 and H3, we use a categorical variable identifying the ultimate selectorate in parties’ candidate selection process, according to formal rules in party statutes. The PPDB data identifies which part of the party has the final say in deciding or approving the final list of candidates. The data allows us to identify four categories: the party members or local organisations (18 parties), the regional or national party organisation (42 parties), the party leader (6 parties) and a mixture of local party/members and party organisation (10 parties) ⁷ .

In order to test H4, we use the ‘organisational structure’ component of the assembly-based intra-party democracy measure drawn from the PPDB database and developed by Poguntke et al. (Poguntke et al., 2016; von dem Berge and Poguntke, 2017). The organisational structure component index synthesises the involvement of members in the life of the party independently from their involvement during personnel selection processes (including candidate selection). It covers the question of the frequency of congresses, the people entitled to vote at them, the composition of the party’s executive bodies and their voting rights. The index can range from 0 (no internal democracy) to 1 (high internal democracy). In our dataset, the index ranges from 0.375 (National Liberal Party in Romania) to 0.91 (Green Party of England and Wales), with a mean value of 0.66. ⁸ The index is normally distributed. ⁹

Control variables

Based on previous research (see above), we control for party size ¹⁰ (percentage of seats in parliament), status of the party in opposition (coded 0: 45 parties) or in government (coded 1: 25 parties), and we also include party age in years (min. 1, max.176, mean 52.6), given that a party’s level of resources could be related to its access to power and level of institutionalisation. We then control for party family, in order to control for a potential ‘ideological’ effect. We distinguish seven categories: Radical Left (6 parties), Social Democrats (12), Green (7), Liberals (14), Christian Democrats/Conservatives (19), Radical Right (8) and an ‘Other’ category (4), including mostly (ethno)regionalist parties but also parties like the Pirate Party in Germany.

At the system level, we first control for the type of electoral system. We consider the type of formula (party list proportional or single-member plurality system) and, for list systems, the openness of the ballot (i.e. whether voters can cast preferences for individual candidates and/or alter the order of the list). We employ four categories, from the most party-centred to the most candidate-centred: closed-list systems (Norway, Romania, Italy ¹¹ ), (semi)open list systems (Belgium, the Czech Republic, Greece, Finland, Sweden), mixed systems (Germany and Hungary) and single-member plurality (United Kingdom). A second system-level control variable relates to the age of the democratic system, as younger democracies can display more political instability, resulting for instance in more frequent party switching (Klein, 2021). We use a simple dichotomous variable distinguishing between ‘young’ (i.e. Czech Republic, Greece, Hungary, Romania) and ‘old’ democracies. ¹²

Given that the literature has found effects of party and career socialisation on the tendency of (elected) candidates to adopt a party-delegate style, we should also take into account these factors. However, there is missing data across the dataset regarding candidates’ previous experience at national and local levels (in party or in political office), as well as regarding length of membership. For each set of these variables, data was missing for one, two or even three countries – which then affected substantially the sample size. We nevertheless provide robustness checks in the Appendix (see Table 3), by testing the effects of two variables: average membership length of the party’s candidates, and average previous experience of the party’s candidates in local politics. Including these controls does not alter the findings.